There are now more than 30 millisecond pulsars known to be associated with the Galactic disk. The majority of these have been discovered in just the last few years as the result of large-scale all-sky surveys. The properties of the population vary tremendously. One unique object hosts a planetary system, more than half of those discovered possess white dwarf companions, two have extremely low-mass companions that are undergoing mass-loss and several others appear to be solitary. In this review I discuss the methods employed to find these millisecond pulsars, the parallels with early surveys for “normal” pulsars, and possible strategies for future searches.

We have surveyed 15,900 deg2 of the northern sky for millisecond pulsars at 370 MHz with the Green Bank 140ft telescope. The telescope was driven along lines of constant declination such that a given point on the sky was observed for two minutes. A Fourier transform spectrometer synthesized 512-channel spectra across 40 MHz in each of two polarizations. Total intensity levels were boxcar averaged at intervals of 256 μs, quantized to 1 bit, and written to tape. Data were processed with the Cray C90 of the Pittsburgh Supercomputer Center. The flux limit of the survey was 8 mJy for slow pulsars. A further 1500 deg2 were observed with 20 MHz bandwidth, yielding 71% the usual sensitivity.

Eighty-four slow pulsars were detected, of which six were previously unknown. Three recycled pulsars were detected: the relativistic binary B1534+12; the neutron star-white dwarf binary J1022+1001, discovered nearly simultaneously at Arecibo; and J1518+4904, a previously unknown pulsar with a 41 ms rotation period.

The majority of known pulsars have been discovered by pulsar searches at low radio frequencies (v < 1 GHz). However, such searches are subject to various deleterious effects, viz the Galactic background radiation (∝ v−2.8), dispersion smearing (∝ v−3) and also scatter broadening (∝ v−4.4). Dispersion smearing and, in particular, scatter broadening prohibit the detection of pulsars with high dispersion measures at low frequencies (cf. Fig. 1a). This is highlighted by the fact that all 11 known pulsars with DM>600 cm−3pc have been discovered during the only two surveys performed to date above 1 GHz, i.e. at 1.4 GHz by Clifton et al. (1992) and at 1.5 GHz by Johnston et al. (1992). However, scattering is still a limiting factor at even 1.4/1.5 GHz. For example B1750—24 is observed with a double component profile at 4.85 GHz (Kijak et al. 1996), whereas at 1.4 GHz the components are completely smeared out due to scatter broadening (cf. Clifton et al. 1992). Therefore, the galactic population of highly dispersed pulsars is still not known. In order to reveal this hidden sample, we have recently started a search in Effelsberg at 4.85 GHz where limitations due to scattering are essentially not existent (see Fig. 1a). The use of this extraordinary high frequency for pulsar searches enables us to observe with a large bandwidth but a small number of filterbank channels, so that the necessary computer power is radically reduced. However, the general steepness of pulsar spectra demands a highly sensitive observing system, otherwise, only the most luminous sources can be detected. A serious disadvantage of a high frequency search is the small telescope beam requiring a lot of observing time to search even a small area of the sky. A restriction of the search area is therefore highly recommended.

The Westerbork Northern Sky Survey is an imaging survey carried out at 325 MHz by the Westerbork Synthesis Array Telescope (de Bruyn, 1996). It covers the northern sky above declination 30° with a resolution of 54” and a limiting flux density (5σ) of circa 15 mJy/beam. The WENSS also provides polarization maps of this part of the sky. About one fourth of the WENSS-325 area will also be covered at 608 MHz to the same limiting flux density. Table 1 gives the parameters of these surveys. It also contains the parameters of the FIRST (Faint Images of the Radio Sky at Twenty-cm, Becker et al. 1995), which is a high resolution survey carried out at 1500 MHz.

The Westerbork Synthesis Radio Telescope (WSRT) has recently been fitted with a new backend designed for pulsar detection. This backend is a filterbank with 32 channels, for both the X and Y polarised signals. A novel feature is that the channel bandwidths are digitally tunable between 6.25 kHz and 2.8 MHz (Vasisht et al. in preparation). The time constant of the integration which follows detection can be digitally set at values between circa 50 ms and 10 μS. For each observing frequency and each pulsar (with a known dispersion measure) we can choose an optimal bandwidth and signal smoothing.

The Following design of a Pulsar Survey System with a Huge Interferometric Array is discussed. (1) A Huge Array of 320m × 320m with 256(= 16 × 16) spherical dishes of 20m in diameter, (2) RF is 1.4 GHz, (3) Receivers are frequency modified GPS(1.57542, 1.2276GHz) ones, (4) Phase and Delay Tracking, (5) Phase Calibration using Differential GPS(GPS signal × 2 → CW), (6) (2+1+2)D FFT /256 pixel Imaging + 256 ch Filtering + Dispersion Removal and Period finding(17M point floating 2DFFT by DSP), (8) Observing time required for the same sensitivity to Arecibo τ/τA: 0.88(Timing), 0.003(Survey).

The phase-coherent recording of pulsar data and subsequent software dispersion removal provide a flexible way to reach the limits of high time resolution, useful for more precise pulse timing and the study of fast signal fluctuations within a pulse. Because of the huge data rate and lack of adequate recording and computing capabilities, this technique has been used mostly only for small pulsar data sets. In recent years, however, the development of very capable, reasonably inexpensive high-speed recording systems and computers has made feasible the notion of pulsar baseband recording and subsequent processing with a workstation/computer. In this paper we discuss the development of a phase-coherent baseband processing system for radio pulsar observations. This system is based on the S2 VLBI recorder developed at ISTS/York University in Toronto, Canada. We present preliminary first results for data from the Vela pulsar, obtained at Parkes, Australia, and processed at ISTS/York University, and discuss plans for future developments.

We have built a new radio astronomical receiving system designed specifically for very high precision timing and polarimetry of fast pulsars. Unlike most detectors currently used to study pulsars, this intrument does not square the received signal at the time of observation. Instead, voltages proportional to the instantaneous electric vectors of incoming signals are digitized, time-tagged, and recorded on high speed magnetic media. We have tested the system using a 5 MHz bandwidth signal with 2-bit digitization at the Greenbank 140 foot telescope. Full polarization information was obtained with a 0.2 μs time resolution. We have used this system to study the giant pulses emitted by PSR B0531+21 and PSR B1937+21, to determine high precision dispersion measures, and to perform high precision timing and polarimetry.

The Navy/Berkeley Pulsar Processor (NBPP) is a digital filterbank which is based on the successful Digital Filter Boards constructed for the University of California Coherent Dispersion Removal Processor (CDRP). One 9U VME crate holds a complete system with 96 channels × 2 polarizations. Running at maximum speed, the full crate will cover 1.8 MHz/channel = 172 MHz total bandwidth. Narrower bandwidths are available by running the system clock slower. The digitization has 4 bits of precision for good dynamic range and minimal quantization effects. The output data has variable time resolution; typically 50-100 μs for search mode, and 2.4 μs for timing mode. A flexible analog mixer/filter front end provides variable width filters, programmable attenuators, and input power level measurements, along with IF-to-baseband mixing and amplification capability. Data acquisition software running on a Sun Sparcstation-20 is based on software developed for controlling the Penn State Pulsar Machine (PSPM).

Communications Research Laboratory (CRL) developed a new millisecond pulsar observation system which used an Acousto-Optic Spectrometer (A0S). An AOS is simple and effective at dividing a wide band signal into many channels simultaneously. Using this system, we started preliminary observations of PSR1937+21 for the purpose of precise measurements of millisecond pulsar’s pulse timing.

Fig. 1 shows a block diagram of our observation system using the 34m antenna of CRL. This antenna has several receivers (1.5, 2, 5, 8, 10, 15, 22, 43 GHz) with selective polarizer. We use 1.5GHz or 2GHz band for pulsar observations. An IF signal with 50MHz bandwidth is divided to 200kHz × 256ch by the AOS. The speed of data transporting rate from the camera is about 13μs which limits a time resolution. We get 100 bins per one period for PSR1937+21. The video averaging processor works as two 8bits A/D converters and an averager which allows 224 pulses’ addition in each channel. The averaged data of each channel are combined after de-dispersion processing which is carried out in 1/1000 steps of the pulsar period. ¿From this final profile, the peak phase is defined as arrival pulse timing. Host#2 calculates the a priori pulse period by TEMPO program, and supplies it in real time to the synthesizer which controls the averaging trigger clock of the timing signal generator. The reference clock of this system is synchronized with UTC (CRL) via the GPS satellites. The difference between the reference clock and the internal clock of the timing signal generator is monitored using a time interval counter.

The recent shift of opinion concerning the distribution of pulsar velocities at birth with a mean value of 400-500 km/s has several ramifications for our understanding of the origin and evolution of pulsars. From the present data, there is no convincing evidence for a causal correlation between velocity and magnetic field, any magnetic field evolution in neutron stars or that galactic neutron stars once active as radio pulsars are the sources of γ-ray bursts. Asymmetric kicks seem the most plausible origin for the high velocities.

One of the important unsolved problems in pulsar astrophysics concerns the formation and evolution of their magnetic fields. We summarize measurements of braking indices and their implications for the spin and magnetic field evolution of young pulsars. An analysis of the period-period derivative diagram suggests one or more of the following: (a) a substantial population of young pulsars remains undiscovered; (b) a large fraction of all slow pulsars may be recycled; (c) magnetic fields in isolated pulsars decay by a factor of a few in the first few million years. We also note that the observed P-Ṗ distribution of millisecond pulsars may be shaped by luminosity evolution.

We perform radio pulsar population synthesis to study the evolution of the pulsar population. In doing this, we continue our earlier work on this subject (Bhattacharya et al. 1992). We have extended our work by 1) calculating orbits in the whole galaxy (the simulation of the observations however is still limited to the solar neighborhood), 2) using the high birth velocities of radio pulsars from Lyne & Lorimer (1994) and 3) implementing the new model by Taylor & Cordes (1993) of the distribution of free electrons in the galaxy. A full report of this work will be given in Hartman et al. (1996).

Monte Carlo simulations of the evolution of pulsars are carried out in order to compare with the recent measurement of the pulsar transverse velocity by Lyne & Lorimer (1994). The new electron density distribution model of Taylor & Cordes (1993) is adopted in the simulation. Accurate pulsar orbits in the Galactic gravitational field are calculated. It is found that the constant magnetic field model of pulsars can account for the new measurement of the pulsar transverse velocity and the apparent correlation between the strength of the magnetic field and the transverse velocity of the pulsars. The present finding confirms the validity of the constant magnetic field model of pulsars and consolidates the idea that the apparent correlation between the strength of the magnetic field and the transverse velocity of the pulsars is caused by observational selection effects.

We have developed a computer code (Tauris & Bailes 1996) to follow the evolution of a binary system from the zero-age main sequence to its “final” state as a binary millisecond pulsar (BMSP), at all stages keeping careful track of the mass and orbital separation of the two stars.

To help determine the origin of millisecond pulsars, we compute the space velocities predicted by various models of their formation. It is difficult to produce a millisecond pulsar velocity greater than 270 km s−1 with any model, unless the formation of the neutron star is accompanied by some form of asymmetric kick. We obtain average 3-D system velocities of 〈vrecoil〉= 99.6, 137.6 and 160.7 km s−1 using Gaussian kicks of 〈vkick〉=0, 200 and 450 km s−1 (σ=0, 100 and 200 km s - 1, respectively). Our computations show that, in general, we expect those systems with shorter orbital periods to have larger velocities than those with longer periods, but any relation between the final orbital period and space velocity is fairly weak, especially if asymmetries are involved.

Lyne & Lorimer (1994) argue that the velocities of young pulsars are much higher than was previously thought. However, recent radio pulsar population synthesis work by Hartman et al. (1996, these proceedings) shows that simulations that use a birth velocity distribution with more pulsars at low velocities, Phinney’s modification of the Paczyński (1990) distribution

describes the observed magnetic field, period and luminosity distribution equally well as simulations that use the Lyne & Lorimer distribution.

We use this radio pulsar population synthesis model to obtain the proper motion distribution of the simulated sample, and from that we calculate the transverse velocity distribution to compare it with the observed transverse velocity distribution. A full report of this work will be given in Hartman (1996).

We summarize our efforts at understanding the galactic neutron star (NS) population using likelihood and bayesian analyses. These include determination of the velocity distribution of young, high-magnetic field objects; the spatial, velocity, period and luminosity distributions of millisecond pulsars; and a full analysis of high-field objects including modeling of pulsars’ radio beams with the implicit assumption that such objects, for fixed period and period derivative, are standard candles.